Sample collection
The living adults of ascidian (H. roretzi) were collected in four distinct months (January, April, July, and October 2018) that briefly represent four main seasons. No apparent morphological changes among animals from different seasons were observed (Fig. 1A and B). Usually the peritrophic membranes of ascidians formed long stringy shape twist filled with dark fecal materials (red arrow, Fig. 1C). However, the peritrophic membranes of the gut became lighter and slimmer, and were covered with sticky secretions without food supply for 2 days and longer (Fig. 1D-F).
Stool samples were collected to delineate the changes of gut microbiota by season, before and after starvation, using 16S rRNA gene amplicon sequencing (Fig. 1G). In order to further understand host-microbe interaction, the gut microbiome in Winter season (January 2018) were isolated for metabolite profiling. Meanwhile, stool samples and ascidian peritrophic tissue samples before (day 0) and after starvation (day 2, 4, and 6) in Winter season were conducted with shotgun metagenomic and transcriptomic sequencing for bacterial and host gene metabolic functional analysis, respectively (Fig. 1G).
Ascidian gut microbiota compared with that of marine environment
We first used 16S rRNA gene hypervariable V4 region amplicon sequencing to compare the difference of microbial communities between the gut and the marine environments. Four seawater samples in each season (n = 16) and five stool samples at each day timepoint of starvation (n = 80) were surveyed, with a total of 4,813,906 high-quality sequences generated from 96 samples (mean ± s.d. of 50,144 ± 9,682). A rarefaction analysis of 20,000 reads per sample clustered short reads into 20,992 amplicon sequence variants (ASVs) that represented 54 bacterial phyla. Among them, 16 phyla were detectable at ³1% relative abundance in at least one sample (Table S1). Proteobacteria (mean relative abundance of 61.1%) was the most predominant bacterial phylum in the surveyed samples, followed by Bacteroidetes (11.2%) and Firmicutes (6.5%) (Fig. 2A).
As expected, we observed differential bacterial communities between samples from seawater and ascidian peritrophic membranes, as discriminated by a principal coordinate analysis (PCoA) using either UniFrac distances or Bray-Curtis dissimilarities (Fig. 2B and Fig. S1). A permutational multivariate analysis of variance (PERMANOVA) using the adonis2 function in R’s package ‘vegan’ based on unweighted UniFrac distances (mean distance between seawater and stool = 0.0531; p<0.001) found a more distinct discrimination in microbial community composition when compared to the weighted UniFrac distances (0.0497; p = 0.003) (Figure S1), indicating that the clustering between ascidian gut and marine seawater samples was driven more by the presence/absence of bacterial ASVs (unweighted) rather than the proportion of microbial community members (weighted). For example, a significant increase of the relative abundance of Bacteroidetes and Epsilonbacteraeota were observed in seawater (Fig. 2C, Table S1) whereas Firmicutes was more common in ascidian stool samples (Fig. 2C). When ASVs were summarized at the order levels, Flavobacteriales, Oceanospirillales, Alteromonadales, and Campylobacterales were largely observed in seawater (mean relative abundance > 5%, MWU p < 0.002), while ascidian stool samples were mainly dominated by Xanthomonadales, Rhizobiales, Legionellales, and Bacteroidales (Table S2), indicating that the bacterial communities may form the strong niche adaptation. In line with differential compositions and abundances, the microbial community of ascidian stool samples showed higher alpha diversities when compared to the seawater (Fig. 2D and Figure S2).
Ascidian gut microbiota changed by season and starvation stress
In order to elucidate the changes of ascidian gut microbiota by season and starvation stress, we refined the ASV table by excluding the seawater samples. Overall, ascidian gut microbiota was mainly dominated by Proteobacteria (mean relative abundance of 46%, represented by Rhodobacterales, Xanthomonadales, Rhizobiales, and Legionellales), followed by Bacteroidetes (8%, represented by Bacteroidales) and Firmicutes (5%, represented by Clostridiales) (Table S3). A PERMANOVA test using Bray-Curtis dissimilarities based on the ASV table indicated that approximately 54% of variation in microbial community composition could be attributed to season (Df = 3, R2 = 0.359, pseudo F = 18.843, p <0.001), starvation (Df = 1, R2 = 0.080, pseudo F = 12.609, p <0.001) and the combination of season and starvation (Df = 3, R2 = 0.103, pseudo F = 5.384, p <0.001), which was supported by the PCoA analysis that the majority of microbial variability was associated with differences between seasons (Fig. 3A). Similarly, we found significant changes of the alpha diversities of gut microbial communities across season (Fig. 3B) and starvation (Fig. 3C).
The relative abundance analysis of bacterial orders revealed that ascidian gut microbiota presented seasonal variation (Fig. 3D and Figure S3, Table S3). For example, Rhizobiales was highly abundant in stool samples collected in January but rarely observed in other seasons (Fig. 4A). Babeliales, Vibrionales, and Xanthomonadales seemed to uniquely form dominant population in April, July, and October, respectively (Fig. 4A). In contrast, the colonization of some bacterial orders might be season-specific. For example, stool samples collected in January and October contained extremely low proportion of Clostridiales and Microtrichales, respectively (Fig. 4A). Bacteroidales and Saccharimonadales were rarely found in Jan/Apr and Jul/Oct, respectively. Interestingly, Xanthomonadales was commonly found in both ascidian stool samples (46.2% vs 0.1%, p<0.001) and seawater (6.7% vs 0.1%, p<0.001) collected in October but not in other seasons, implying that gut bacterial transmission from marine environment is possible (Figure S4, Table S4).
Consistent with the decreased alpha diversity of gut microbiota during starvation (Fig. 3C), a number of microbes largely changed in the relative abundances (Figure S5, Table S3). We found 13 bacterial orders prevalently decreased across starvation while another 11 becoming more resistant, with statistical significance in at least one season. As shown in Fig. 4B, for example, Synechococcales and Pirellulales, two predominant gut bacterial orders in aquafarm condition in most of seasons, were dramatically depressed when food and nutrition elements were lacking (mean relative abundance of 9.9% vs 0.4%, q <0.001; 4.7% vs 0.9%, q <0.001). In contrast, some rare bacteria in certain seasons, such as Xanthomonadales, Legionellales, Alteromonadales, and Corynebacteriales, became booming in starvation condition. The relative abundance analysis of bacterial genus also revealed similar seasonal variation (Table S5).
Functional profile of ascidian gut microbiota based on 16S rRNA gene amplicon sequencing
Differential gut microbial communities observed between habitats, seasons, and starvation conditions indicates that these factors may enrich for functionally different microbial communities. Hence, we used PICRUSt2 (Phylogenetic Investigation of Communities by Reconstruction of Unobserved States) to predict functional pathways based on the composition of the microbial communities and produced Kyoto Encyclopaedia of Genes and Genomes (KEGG) Orthology (KO) abundance profiles. Results of the summarized KO pathways were supported by spare partial least squares discriminant analysis (sPLSDA) using the first three ordination components that show clustering of samples mainly by seasons and habitats (Figure S6).
Next, we attempted to identify the metabolic functions that discriminated the ascidian gut microbial communities before and after starvation (Fig. 5A). As shown in Table S6, we found 26 up- and 22 down-regulated pathways across starvation, with statistical significance in one season and more. Among them, the functions involving photosynthesis (ko00195, ko00196) and its related biosynthesis (ko00710, ko00906) were dramatically depressed (baseMean > 1000, |log2FoldChange| > 1, q<0.001) (Fig. 5A), probably a result of the reduced colonization of Synechococcales in the starvation condition (Fig. 4B). In contrast, enrichments of Xanthomonadales and Legionellales in relative abundances after starvation might facilitate bile acid biosynthesis (ko00120, ko00121) (Fig. 4B and Fig. 5B). Xanthomonadales and Corynebacteriales might also contribute linoleic acid metabolism (ko00591) and biosynthesis of siderophore group nonribosomal peptides (ko01053). The moderately increased metabolism pathway involving bacterial secretion system (ko03070, baseMean = 25600, log2FoldChange = 0.36, q = 0.001) might explain in part the observation of sticky secretions covering the surface of ascidian peritrophic membranes during the starvation (Table S6). It is worth noting, however, that the limited resolution of partial 16S rRNA gene in discriminating bacterial phylotypes, as well as a possible lack of marine animal PICRUST2 reference microbial genomes may have limited resolution of functional prediction, given the relatively high scores of the weighted Nearest Sequenced Taxon Index (0.17 ± 0.10).
Metabolic changes of gut microbiome and host across starvation
In order to further understand the host-microbe interaction, the ascidian stool samples and peritrophic tissues collected in January were conducted for metabolic profiling using high-performance liquid chromatography. Among 37,538 identified metabolites, 1,157 of them could be annotated as known ones using mass spectrometry data (MS2 spectrum) and metabolic reaction network (MRN)-based recursive algorithm (MetDNA) (Table S7). The PCoA analysis based on the abundance of all the identified metabolites clearly discriminated stool samples from the ascidian tissues (Fig. 6A), implying differential metabolic profiles between microbiota and host. We also observed distinct separation of stool samples before (Day 0) and after starvation (Day 246). However, the metabolic profiles of the ascidian tissues did not significantly change before and after starvation, suggesting that starvation mainly has significant impact on the gut microbiome rather than host (Fig. 6B, Table S8). When the abundances of metabolites were visualized in a heatmap, we observed a pattern of metabolites highly expressed in stool samples across starvation when compared with those in aquafarm condition (see green rectangle in Fig. 6C), such as the pathways involving linolenic acid metabolism, methane metabolism, and cyanoamino acid metabolism (Fig. 6D). In contrast, a number of abundant metabolites in aquafarm condition were dramatically depressed (see red rectangle in Fig. 6C), such as phenylalanine metabolism, phenylalanine tyrosine, tryptophan biosynthesis, and D-glutamine and D-glutamate metabolism (Fig. 6D). Some metabolites might be host- or bacteria-specific. For example, linoleic acid, a product from plants and green algae, could also been synthesized by bacteria [30][31, 32]. Interestingly, there was limited impact of starvation in regulating metabolites of host tissue samples implying that the dysbiosis of gut microbiome may be mainly responsible for the changes of metabolites across starvation.
Contribution of gut microbiome in metabolite changes
To determine the influence of gut microbiome in changing metabolic pathways before and after starvation, we first performed the correlation analysis of metabolites, and found that the abundance between bacteria and metabolites were highly correlated. (Figure S7), suggesting bacterial origin of metabolites. For example, phosphatidylcholine lyso and arachidonate in arachidonic acid metabolism pathway were highly correlative with Rhodobacteriales, Flavobacteriales, vibrionales, and Spirochaetales etc. (Figure S7).
To further reveal the origin and the difference of metabolites between ascidian gut microbiome and peritrophic tissue, transcriptome sequencing for tissue samples (n = 4) (Table S9) and metagenomic sequencing for stool samples (n = 4) (Table S10) in winter, before and after starvation, were performed, respectively. A total of 176.6 and 242.3 million short reads were obtained from transcriptome and metagenome sequencing, respectively. One hundred and eighteen KEGG pathways were shared by both stool and tissue samples across metabolome and metagenomic analyses (Figure S8A). Interestingly, eighty-three pathways (9 + 66 + 8) observed in either stool or tissue samples with metabolome analysis were detectable in the metagenomic annotation of gut microbiome but not peritrophic tissue, indicating that the gut microbiome largely contribute the synthesis and decomposition of metabolites, such as alox15 and beta-carotene 3-hydroxylase (Figure S8B and S8C).
Potential metabolic pathways involving the interactions between ascidian gut microbiome and host were proposed. As observed, the pigment compounds (such as astaxanthin and Xanthophyll), plant-like polyunsaturated fatty acids and esters, hormone signal substance, plant hormones (such as salicylic acid and stearidonic acid), C18 unsaturated fatty acids (such as oleic acid, linoleic acid, and linolenic), phenylalanine, benzoate, salicylic acid, and stearidonic acid were significantly enriched in the gut (Fig. 7A). These bacterial origin metabolites were likely absorbed and played crucial roles on host energy supports, inflammation balancing, and body defense through glucose and lipid metabolism pathways. For example, plant hormones and C18 unsaturated fatty acids are common signaling substances constituting the systemic acquired resistance (SAR) immune system in ascidian gut (Fig.7A). In contrast, unsaturated fatty acid-related metabolism including arachidonic acid and linoleic acid was significantly enhanced at Day 246, suggesting that the gut microbiome may serve as nutritional supplements involved in body functions when ascidian is under the starvation stress (Fig. 7A). A number of gut bacterial orders were deduced as metabolic contributors (Fig. 7B). For example, Rhodobacterales and Xanthomondadales may produce carnitine, cholic acid (CA), and branched-chain-amino-acids (BCAA) that regulate glucose and lipid metabolism for energy maintenance; Solirubrobacterales and Rhodobacterales could be the source bacteria related to inflammation balancing and systemic immunity (Fig. 7B).